• When working with data, there are often multiple ways to store the exact same information
• For example, we can sometimes store data in what is called “wide” or “long” format

• Sometimes we have other problems with our data
• For example, we may have some incorrect/duplicated observations, redundant variables, and missing values
• These are usually less difficult to deal with in traditional software, but can also be a pain

• Tidy data is a concept developed by Hadley Wickham
• The goal of tidy data is to have one data format which is simple, easy to work with, and can be used for any analysis

• Tidy data is data in which:
  • Each variable has its own column
  • Each observation has its own row
  • Each value has its own cell

• Observational unit is whatever we measure (e.g. subjects, cases, trials, schools, etc…)
• Each observational unit has ALL the IVs and the DV that we’re interested in
• If we measure a subject repeatedly (e.g. in repeated measure design, over time or in different conditions), then each measure is an observational unit . In other words, each unique value of the DV should be its own observational unit
• You know your data is tidy when each row of your data is exactly one observational unit, each column is either IV or DV (or some auxilary variable, e.g. ID), and the DV is in a single column

• dplyr is an R package developed by Hadley Wickham
• dplyr provides a set of verbs for manipulating dataframes:
  • select: select or drop variables
  • filter: filter observations based on conditions
  • arrange : order observations by values in variables
  • mutate: modify existing variables or create new ones
  • gather : gather a variable that is spread across multiple columns
  • spread : spread mutliple variables that are squeezed in one column
  • group_by and summarize: create cross-tabulations using variables

• The pipe (“%>%”) is a special thing that makes R great
• R is a functional programming language, meaning that we pass objects onto functions to make things happen. For example, if we run the code mean(c(1, 3, 5, 7), we take the mean of the numbers 1, 3, 5, and 7 (= 4)
• Pipe allows us to invert the relationship, which makes for a more readable code:

c(1, 3, 5, 7) %>% # First, take a list of number 1, 3, 5, 7, THEN
  mean() # Take its mean

## [1] 4

• Pipe makes the code much more readable for humans!

• For example, if are were only interested in the ID, age and smoker variables:

my_data %>%
  select(ID, age, smoker) %>%
  knitr::kable()

• If we want to drop a variable from a dataset, we simple add the “-”. For example, if we want to drop the ID variable:

my_data %>%
  select(- ID) %>%
  knitr::kable()